Tuesday, January 12, 2010 ... Français/Deutsch/Español/Česky/Japanese/Related posts from blogosphere

IPCC types read Lindzen-Choi 2009

A few months ago, Richard Lindzen and Yong-Sang Choi, his postdoc, published an interesting paper claiming to have identified a serious discrepancy between the reality and models when it comes to the response of the energy flows in the tropics to a changed temperature. The main result was that in reality, the feedback coefficient seems to be negative and the resulting climate sensitivity seems to be below 1 degree Celsius:

On the determination of climate feedbacks from ERBE data (PDF)
Here is the most famous picture from the paper:



I have had a specific problem with the paper. But let me enumerate a couple of links, starting with some old ones and continuing with the newest ones:
TRF: Spencer on Lindzen-Choi (his WWW)
TRF: Climate feedbacks from measured energy flows
TRF: Lord Monckton promotes it on Glenn Beck

Revkin: a rebuttal to a cool paper by LC
RC: Trenberth et al. respond (technically)
RC: Trenberth et al. (link to their paper)
RC: Gavin Schmidt about finding reviewers of LC
The paper written to find errors in Lindzen-Choi 2009 is by Trenberth, Fasullo, O'Dell, and Wong, TFOW 2010. There exists a subtle non-uniformity in this group of 4 authors. The last one has communicated the results to Richard while the remaining three have written a post for Real Climate ;-). Takmeng Wong is not a professional alarmist but rather a member of the experimental teams (CERES, ERBE), located at NASA Langley.




Gavin Schmidt and others in his team have "relatively" praised the Lindzen-Choi paper: Schmidt has even dared to say that it wasn't "obvious nonsense". And Richard has been nice about the Trenberth et al. results, too:
Thanks [to Revkin] for passing on the paper [by Trenberth et al.]. I had not seen it. However, Wong had communicated much of the material independently, and some of it is certainly valid. However, we have addressed the criticisms and have shown that the results remain — especially the profound disconnect between models and observations. We are currently preparing a new version, and it should be ready shortly.
I have already seen the new version and it looks really fine to me so far but the rest of this text was written before I had seen it.

The rest of the Real Climate article about the "not obvious nonsense" describes the creation of the TFOW team and Trenberth's sorrow that his original attempt to shoot Lindzen-Choi down was instantly rejected. Oh, the poor alarmists are surely so discriminated against by the reviewers and funding agencies!

And the reviews of Lindzen-Choi were so extremely favorable - it must have been a gross injustice, right? ;-) Gaia is already going after the reviewers' neck.

The three non-Wong TFO[W] authors wrote a technical summary for Real Climate. It describes five main points.

Missing black-body radiation

The last one coincides with mine. Trenberth et al. complain that the black body function is omitted in the list of forcings. When corrected, the feedback coefficient and doubling sensitivity change from -1.1 and 0.5 °C to -0.125 and 0.82 °C. Note that this change is not really "qualitative" because the feedback remains slightly negative and the expected warming is unspectacular.

That's probably why this point - arguably the most crisp one - wasn't written at the beginning of the Trenberth et al. article: the readers could see that even when the most "identifiable" problem is fixed, the warming remains negligible.

In this last point, they also say that the tropics shouldn't be studied separately because they're not a closed system. I feel uneasy about this point, too. After all, even if one wrote a completely correct paper of this type, he could only determine the sensitivity (warming) as seen in the tropics which may importantly differ from the value in other regions.

Arguments remain across the alarmist and realist communities on whether or not the black-body term should be added, and if it should, whether its absence is just a problem of the models; models as well as Lindzen-Choi phenonmenological methodology; or a real "paradoxical behavior" in the observed data.

Robustness

There are four other points of criticism - Trenberth et al. surely try to escalate the criticism to partisan levels that are not backed by the strength of their actual arguments. In the first point, they say that Lindzen and Choi are not "robust". TFOW 2010 claim that the calculated sensitivity dramatically depends on the choice of the end points where the responses are calculated - even on changes by one month - and that they have an "objective" method to choose the "right" endpoints.

Well, it could also be "objectively" designed to get the most convenient result that is possible. Let me say that the relatively clear trends seen in the 12 pictures above make it hard for me to believe that one can radically change the outcome. One may perhaps increase the noise but I do think that the 12 graphs do show some "signal" and TFOW 2010 may only try to obscure it.

Air-sea interactions in the tropics

This looks like some kind of "role reversal" to me because what TFOW 2010 write here is the kind of skeptical arguments that I don't enjoy too much - "everything is complex so any quantitative argument should be ignored". In particular, Lindzen and Choi assume that the sea surface temperatures have an impact on the clouds.

On the other hand, TFOW claim that El Nino/La Nina dynamics has much bigger an impact than the sea surface temperature itself because everything is interconnected, complex, Gaia, global, and so on, and can't be measured locally (TFOW argue with rain, wind, telekinesis etc.).

I find these comments about "diversity of El Nino effects vs sea surface temperatures" strange because the presence of El Nino conditions themselves is defined by the sea surface temperature anomalies in the equatorial Pacific Ocean. And the very logic of ENSO is that one assumes that the temperature, wind, and precipitation patterns are correlated with it, too. So the ENSO effects do impact the Lindzen-Choi analysis - because they should. In some sense, it is about them.

Moveover, it seems clear that if there were an important effect on the clouds that is "completely independent" from the sea surface temperature, it would cause much less robust trends in the 12 linear regressions.

"More robust" methods: all is OK

TFOW claim that there are "more robust" methods that show no discrepancy and Lindzen-Choi should therefore be dismissed. This is a bizarre comment, at a very philosophical level. It's enough to find one - or a few - observations that contradict a hypothesis with a sufficient confidence, and the theory is dead. You may avoid the inconvenient tests but they won't cease to exist because of that.

TFOW haven't explained why they think that these "all is OK tests" are more robust than Lindzen and Choi. So I think it is sensible to think that the "all is OK" tests were more likely to have been published than tests that actually do isolate some discrepancies between models and the real climate (such tests are surely not fashionable), for reasons that all of us know. At any rate, Lindzen and Choi is now published, too, so it's harder to pretend that it doesn't exist.

Volcanos

TFOW report that some of the models have omitted forcings such as the huge Mount Pinatubo eruption in June 1991 while others have counted it. Well, we still have some models that did take the eruption into account and a disagreement still exists for them.

On the other hand, I don't quite understand why the omission by the other models should necessarily be considered a fault of Lindzen and Choi. What really matters is whether models agree with the reality - and volcano eruptions are a part of this reality.

Summary

I do think that the paper by Lindzen and Choi has some bugs - and I hope they will be mostly corrected in the near future. But my prelimiinary calculations show that the sensitivity remains close to 1 degree Celsius - "the truth is somewhere in between" - when the things are corrected.

And that's a completely unremarkable CO2-induced temperature change especially because it's not far from the effect of one or two volcano eruptions - phenomena that many models are freely omitting. The very fact that it looks "sensible" for some people to build the whole models around CO2 while ignoring effects that are equally or more important - as Trenberth et al. have claimed themselves - shows a bias, a distorted focus of the science driven by the purpose.

Even if the CO2 climate sensitivity were above 2 °C, which I find unlikely, it would still be just one, relatively simple term in all these equations. The bulk of atmospheric science would still have to focus on all the other phenomena - because they're more complex, there's a lot of unknowns about them, and they still matter - and it's just very bad that this opinion is not shared by the overwhelming majority of the climate community today.

Add to del.icio.us Digg this Add to reddit

snail feedback (0) :